Elsevier

Applied Energy

Volume 240, 15 April 2019, Pages 778-792
Applied Energy

A novel locally adaptive method for modeling the spatiotemporal dynamics of global electric power consumption based on DMSP-OLS nighttime stable light data

https://doi.org/10.1016/j.apenergy.2019.02.062Get rights and content

Highlights

  • A novel local-adaptive method was proposed to model global EPC at 1 km resolution.

  • Multiple options were adopted to adaptively correct the NSL data.

  • Various regression models were used to reflect the relationship between EPC and NSL.

  • Our product showed higher spatiotemporal precision compared to the current one.

  • The spatiotemporal dynamics of global electric power consumption were investigated.

Abstract

Timely and reliable estimation of electricity power consumption (EPC) is essential to the rational deployment of electricity power resources. Nighttime stable light (NSL) data from the Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS) have the potential to model global 1-km gridded EPC. A processing chain to estimate EPC includes: (1) NSL data correction; and (2) regression model between EPC statistics and NSL data. For the global gridded EPC estimation, the current approach is to correct the global NSL image in a uniform manner and establish the linear relationships between NSL and EPC. However, the impacts of local socioeconomic inconsistencies on the NSL correction and model establishment are not fully considered. Therefore, in this paper, we propose a novel locally adaptive method for global EPC estimation. Firstly, we set up two options (with or without the correction) for each local area considering the global NSL image is not saturated everywhere. Secondly, three directions (forward, backward, or average) are alternatives for the inter-annual correction to remove the discontinuity effect of NSL data. Thirdly, four optional models (linear, logarithmic, exponential, or second-order polynomial) are adopted for the EPC estimation of each local area with different socioeconomic dynamic. Finally, the options for each step constitute all candidate processing chains, from which the optimal one is adaptively chosen for each local area based on the coefficient of determination. The results demonstrate that our product outperforms the existing one, at global, continental, and national scales. Particularly, the proportion of countries/districts with a high accuracy (MARE (mean of the absolute relative error)  ≤ 10%) increases from 17.8% to 57.8% and the percentage of countries/districts with inaccurate results (MARE > 50%) decreases sharply from 23.0% to 3.7%. This product can enhance the detailed understanding of the spatiotemporal dynamics of global EPC.

Introduction

Along with the tremendous development of the global economy, energy demand has continuously increased over the last century [1], [2]. As an indispensable component of energy, electric power plays a vital role in numerous aspects of modern society, such as improving residential living standards [3], supporting industrial production [4], and promoting commercial transactions [5]. According to the World Bank [6], global electric power consumption (EPC) in 2014 was more than four times higher than that in 1971. In addition to the convenience brought by the massive increase of EPC, the world has also been burdened with accelerated global warming and air pollution due to the accompanying emission of greenhouse gases and other pollutants [7], [8]. Therefore, accurate delineation of the spatiotemporal dynamics of global EPC is a critical prerequisite for investigating both the impacts of EPC and its interaction with the economy and the environment [9], [10].

A wealth of research has investigated the spatiotemporal dynamics of EPC based on the EPC statistics published by related official organizations. For instance, AI-Garni et al. [11] adopted a regression model to forecast EPC in Eastern Saudi Arabia using weather data, global solar radiation, and population as variables. Egelioglu et al. [12] predicted annual EPC by multiple regression analyses of the historical economic databases and EPC statistics for Northern Cyprus. Shiu and Lam [13] examined the causal relationship between EPC and GDP in China by the error-correction model. Huang et al. [14] investigated the electric power supply and demand in China using the Grey-Markov forecasting model. Chujai et al. [15] forecasted the EPC at a household scale with different autoregressive models based on time-series EPC statistics. Cabral et al. [16] developed a spatiotemporal method that considers spatial correlations to predict the EPC in Brazil. These previous studies have been devoted to providing suggestions for governments or organizations. However, for the EPC statistics, the collection process is labor-intensive and time-consuming. Moreover, the EPC statistics are unable to reflect the internal spatial details within the administrative unit, which limits our understanding of the spatiotemporal dynamics of EPC at smaller scales [17], [18]. Compared with the statistics for an entire administrative unit, gridding is a more realistic representation for the investigation of EPC at finer scales. Therefore, efficient methods to produce a spatially gridded representation of worldwide EPC are urgently needed, and it is worth attempting to adopt appropriate spatial gridded data as a proxy for modeling global EPC.

Satellite remotely sensed imagery has been proved to be a reliable way to support large-scale investigations in numerous fields, such as global solar radiation [19], land surface temperature [20], land use and land cover [21], and CO2 emission [22]. The nighttime light (NTL) remote sensing imagery, such as that obtained by the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP-OLS) [23], has the potential for EPC estimation over large areas, because NTL can directly reflect the EPC caused by anthropogenic socio-economic activities at night [24], [25], [26], [27]. Elvidge et al. [28] verified the high log-log relation between the lit areas in DMSP-OLS data and EPC for 200 countries during 1994–1995. Similarly, Lo [29] modeled the logarithmic relationship between DMSP-OLS NTL and EPC for 35 Chinese cities for the year 1997. Amaral et al. [30] found that DMSP-OLS NTL was linearly correlated with the statistical EPC for 1999 in Brazilian Amazonia. Chand et al. [31] analyzed the linear relationship between the increase of EPC and the increase of NTL in the major cities and states of India during 1993–2002. Townsend and Bruce [32] reported a strong second-order polynomial relationship between DMSP-OLS NTL and EPC at the state level in Australia for 1997–2002. Letu et al. [33] estimated EPC in Japan and other Asian countries from saturated-corrected DMSP-OLS data, and found a strong linear correlation [34] between EPC and DMSP-OLS data in Japan. He et al. [35] respectively modeled double-log relationships for different economic regions of the Chinese Mainland from 1995 to 2008 at the county level. Ma et al. [36] attempted three models (linear, power-law, and exponential function) to quantify the relationships between EPC statistics and DMSP-OLS NTL for more than 200 cities in China during 1994–2009, and suggested that the best quantitative model type varies with the different socioeconomic patterns. Xie and Weng [37] explored the country-level relationship between EPC statistics and DMSP-OLS NTL by the logarithmic function. Jing et al. [38] adopted the linear model to correlate EPC with DMSP-OLS NTL data at the provincial level in China. By summarizing the existing literature, it can be found that different types of models have been utilized across different regions when using NTL to estimate EPC, due to the disparity of the social, economic, and urban development status among the different regions.

With respect to the estimation of spatially gridded EPC, Zhao et al. [39] estimated the provincial-level EPC based on the DMSP-OLS and population data in China, and generated pixel-level EPC through disaggregation. Cao et al. [40] proposed a statistics-to-grid scaling down method for mapping gridded EPC in China based on the integration of DMSP-OLS data, population and gross domestic product (GDP). He et al. [41] modeled annual pixel-based EPC in Chinese Mainland with DMSP-OLS and normalized difference vegetation index (NDVI). Xie and Weng [42] estimated gridded EPC of China using DMSP-OLS data, population and enhanced vegetation index (EVI) considering the difference between urban cores and suburban areas. Pan and Li [43] generated 1-km EPC map in China with different vegetation indices and DMSP-OLS data. Most of the existing studies have focused on modeling at the national, regional, or city level, however, research at the global scale is scarce, due to its complexity. An exception and a notable example is the study of Shi et al. [44], where the original NTL images were first corrected worldwide using a uniform framework, and the world was then partitioned into 48 regions according to the geographic locations and socioeconomic levels. Finally, a linear model between the EPC statistics and corrected NTL data for each region was individually built to explore the gridded EPC. Nonetheless, this method does not fully consider the influence of the uniqueness of local socioeconomic development on the following three aspects in the EPC estimation.

  • (1)

    The saturation issue of NTL data: The relatively low radiometric resolution (6 bits) of the OLS sensor results in saturation in the NTL data [45], especially in the centers of large cities. All digital number (DN) values of these saturated pixels are 63, and hence, the disparity within the urban centers cannot be distinguished. In Shi et al. [44], a modified invariant region (MIR) method was globally adopted to reduce the saturated pixels. Nevertheless, saturated pixels are not ubiquitous worldwide, especially in underdeveloped areas. Saturation correction can result in distortion of these unsaturated pixels in suburban and rural areas [46], and reduce the contribution of saturated pixels to the EPC estimation [42]. Therefore, it is not appropriate to globally utilize a unified framework for saturation correction.

  • (2)

    The incomparability and discontinuity effect of NTL data: The original NTL images cannot be directly compared with each other, due to the lack of onboard calibration for the OLS sensor. Specifically, the DN values in the images obtained from the same satellite fluctuate abnormally in different years, and discrepancies occur in the images collected by different satellites for the same year [47]. Shi et al. [44] performed inter-annul correction in a forward direction for the whole world to eliminate the abnormal fluctuation (i.e., discontinuity effect) [43], [44], [48]. However, in addition to the forward direction, other approaches (e.g., backward, average) can also be considered for the correction, resulting in different corrected NTL data [49], [50]. With inappropriately corrected NTL data, the reliability and accuracy of EPC estimation can also be affected. Therefore, it is not reasonable to apply the same approach (e.g., forward) for all the regions with diverse socioeconomic dynamics throughout the world.

  • (3)

    The estimation model between the EPC statistics and the corrected NTL data: Shi et al. [44] employed linear models for all the regions in the world. However, as indicated by the previous studies [35─38], the appropriate type of regression model can vary across areas, owing to the local socioeconomic diversity. Therefore, it is inappropriate to limit the model type to the linear one at the global scale.

To address the aforementioned research questions, we propose a novel locally adaptive method for modeling global EPC. Since the available EPC statistics across the globe are at the national level, the local scale in this study is set as the country/district level. Specifically, for each country/district, two options (with or without correction) are first designed for saturation correction, and three optional directions (forward, backward, or average) are secondly considered for the inter-annual correction. Four alternative models (linear, logarithmic, exponential, or second-order polynomial functions) are then set up to reflect the possible relationships between the EPC and NTL data. Finally, the processing chain composed of the optimal options in the three aspects is adaptively selected for each country/district, to accommodate the local socioeconomic status.

The rest of this paper is organized as follows. Section 2 focuses on the data sources used in this study. Section 3 introduces the proposed locally adaptive selection method for global EPC mapping. The results and discussion are respectively presented in 4 Results, 5 Discussion. Finally, Section 6 sets out our main conclusions.

Section snippets

Data sources

The Version 4 global nighttime stable light (NSL) data of DMSP-OLS for 1992–2013 were obtained from the National Oceanic and Atmospheric Administration-National Geophysical Data Center (NOAA/NGDC) website (http://www.ngdc.noaa.gov/eog/dmsp.html). The NSL images were acquired by six satellites: F10 (1992–1994), F12 (1994–1999), F14 (1997–2003), F15 (2000–2007), F16 (2004–2009), and F18 (2010–2013), covering an area from −180 to 180 degrees in longitude and −65 to 75 degrees in latitude. The 34

Methodology

The proposed locally adaptive method for modeling global EPC consists of four main procedures: (1) decomposition of the global NSL images into national NSL data based on the national boundaries; (2) sequential connection of all possible options in the NSL data correction (including the saturation correction and inter-annual correction) and EPC estimation to form all the candidate processing chains; (3) locally adaptive selection of the optimal processing chains to construct the global EPC; (4)

The saturation correction

To clearly show the effects of the saturation correction on the NSL data for areas with different socioeconomic levels, we visually compared the original NSL with saturation-corrected (VANUI) images for six cities in 2013, using the 30-m resolution Landsat 8 OLI images and the VIIRS/DNB data as the reference to show the urban regions (Fig. 2). Afghanistan, China, and the United States were selected as representative countries, considering their different development levels. Within each country,

Comparison with existing global products

Shi et al. [44] modeled 1-km resolution global EPC maps from 1992 to 2013 by dividing the world into a series of regions, with a linear regression model for each region (hereafter referred to as Shi’s product). Here we compare our estimated global EPC maps with Shi’s product at global, continental, and national levels, respectively, based on the country-level statistics.

Conclusion

The DMSP-OLS nighttime stable light (NSL) images have the ability to model gridded electricity power consumption (EPC) across the globe. However, we need to properly deal with the saturation problem, as well as the incomparability and discontinuity issues existing in the original NSL data, to make the data a reasonable approximation of EPC. The regression model can then be built to quantify the relationship between the EPC statistics and the corrected NSL for the gridded EPC estimation.

Acknowledgements

The research was supported by the National Natural Science Foundation of China under Grant 41771360, the National Program for Support of Top-notch Young Professionals, the Hubei Provincial Natural Science Foundation of China under Grant 2017CFA029, and the National Key R&D Program of China under Grant 2016YFB0501403.

References (68)

  • A. Shiu et al.

    Electricity consumption and economic growth in China

    Energy Policy

    (2004)
  • M. Huang et al.

    Predictive analysis on electric-power supply and demand in China

    Renew Energy

    (2007)
  • W. Huang et al.

    Connecting water and energy: Assessing the impacts of carbon and water constraints on China’s power sector

    Appl Energy

    (2017)
  • J. Qin et al.

    Estimation of monthly-mean daily global solar radiation based on MODIS and TRMM products

    Appl Energy

    (2011)
  • P. Fu et al.

    A time series analysis of urbanization induced land use and land cover change and its impact on land surface temperature with Landsat imagery

    Remote Sens Environ

    (2016)
  • X. Huang et al.

    Multi-level monitoring of subtle urban changes for the megacities of china using high-resolution multi-view satellite imagery

    Remote Sens Environ

    (2017)
  • K. Shi et al.

    Spatiotemporal variations of urban CO2 emissions in China: A multiscale perspective

    Appl Energy

    (2018)
  • K. Shi et al.

    Spatiotemporal variations of CO2 emissions and their impact factors in China: A comparative analysis between the provincial and prefectural levels

    Appl Energy

    (2019)
  • H. Lu et al.

    Spatial effects of carbon dioxide emissions from residential energy consumption: A county-level study using enhanced nocturnal lighting

    Appl Energy

    (2014)
  • H. Xiao et al.

    Spatio-temporal simulation of energy consumption in China's provinces based on satellite night-time light data

    Appl Energy

    (2018)
  • C.D. Elvidge et al.

    Night-time lights of the world: 1994–1995

    ISPRS J Photogram Remote Sens

    (2001)
  • S. Amaral et al.

    Estimating population and energy consumption in Brazilian Amazonia using DMSP night-time satellite data

    Comput Environ Urban Syst

    (2005)
  • T. Ma et al.

    Quantitative estimation of urbanization dynamics using time series of DMSP/OLS nighttime light data: A comparative case study from China's cities

    Remote Sens Environ

    (2012)
  • X. Cao et al.

    Spatialization of electricity consumption of China using saturation-corrected DMSP-OLS data

    Int J Appl Earth Obs Geoinf

    (2014)
  • Y. Xie et al.

    Detecting urban-scale dynamics of electricity consumption at Chinese cities using time-series DMSP-OLS (Defense Meteorological Satellite Program-Operational Linescan System) nighttime light imageries

    Energy

    (2016)
  • K. Shi et al.

    Detecting spatiotemporal dynamics of global electric power consumption using DMSP-OLS nighttime stable light data

    Appl Energy

    (2016)
  • C.D. Elvidge et al.

    Radiance calibration of DMSP-OLS low-light imaging data of human settlements

    Remote Sens Environ

    (1999)
  • J. Wu et al.

    Exploring factors affecting the relationship between light consumption and GDP based on DMSP/OLS nighttime satellite imagery

    Remote Sens Environ

    (2013)
  • D.P. Roy et al.

    Prototyping a global algorithm for systematic fire-affected area mapping using MODIS time series data

    Remote Sens Environ

    (2005)
  • Q. Zhang et al.

    The vegetation adjusted NTL urban index: A new approach to reduce saturation and increase variation in nighttime luminosity

    Remote Sens Environ

    (2013)
  • L. Meng et al.

    Estimating CO2 (carbon dioxide) emissions at urban scales by DMSP/OLS (Defense Meteorological Satellite Program’s Operational Linescan System) nighttime light imagery: methodological challenges and a case study for China

    Energy

    (2014)
  • K. Shi et al.

    Modeling spatiotemporal CO2 (carbon dioxide) emission dynamics in China from DMSP-OLS nighttime stable light data using panel data analysis

    Appl Energy

    (2016)
  • G. Aydin

    Modeling of energy consumption based on economic and demographic factors: The case of Turkey with projections

    Rene Sustain Energy Rev

    (2014)
  • J. Zhao et al.

    Spatio-temporal dynamics of urban residential CO2 emissions and their driving forces in China using the integrated two nighttime light datasets

    Appl Energy

    (2019)
  • Cited by (49)

    • Modeling the spatiotemporal dynamics of global electric power consumption (1992–2019) by utilizing consistent nighttime light data from DMSP-OLS and NPP-VIIRS

      2022, Applied Energy
      Citation Excerpt :

      However, when using time-series VIIRS, the inter-annual discontinuity effect is rarely taken into account, with one exception that Lv et al. [48] performed the same forward continuum correction for VIIRS in China as for DMSP. Nonetheless, the result in [32] has indicated that different inter-annual correction approaches should be applied in regions with diverse socioeconomic status. Therefore, three directions (forward, backward, and average) are considered as options for the inter-annual correction of global VIIRS data in this research.

    View all citing articles on Scopus
    View full text